The group recently attended the DataNinja Spring School, held online from the 23rd to 25th of March. The Spring School was organized by the University of Bielefeld and focused on “Artificial Intelligence – perspectives and challenges”.
Maribel Acosta held a talk on “Symbolic and Sub-symbolic Representations of Knowledge Graphs”. After giving a brief Introduction to Knowledge Graphs, she explained how Knowledge is encoded in a Knowledge Graph in a symbolic way and how reasoning can be performed over those symbolic representations. In the second half of the talk, she covered the topic of Sub-Symbolic representations of Knowledge Graphs, where methods from machine learning are used to embed the Knowledge Graph into a latent space. This representations can then be used to predict missing information in the Knowledge Graphs. She also presented some of her recent work in this field, where she used Restricted Boltzmann Machines to learn those embeddings in an unsupervised way.
Further Talks were held about Deep Reinforcement Learning, AutoML and Explainable AI. The group learned about Explainability in Graph Neural Networks and is now discussing whether those techniques can be used to better explain Knowledge Graph Embeddings.
Link to the Spring School: https://dataninja.nrw/?page_id=762
Paper about Restricted Boltzmann Machines: https://dl.acm.org/doi/pdf/10.1145/3459637.3482377